On 1 May 2004, Dan Kegel <[EMAIL PROTECTED]> wrote: > >Hardcode the socket to ~/.lbfsh/socket. (Well, maybe have an > >option to override.) If the client finds there's no daemon > >listening, it forks one. > > I suppose. Makes me a bit nervous somehow to have daemons > started behind my back, though.
Never-been-done is not necessarily bad. Anyhow, a vaguely similar pattern of an interactive process is used by emacsserver and gconf, and probably others I can't think of at the moment. I think the improvement in friendliness makes up for any nervousness. Now if it were going to accept network connections or something like that, then I would be nervous. But if it works entirely within the security domain of the user who is running it, then I think it's harmless. > If autoforking the daemon was a good idea, the ssh guys would have > had ssh-agent autofork off of ssh, I bet. It might be a better match to run it from ssh-add, since it only needs to exist when you want to remember unlocked keys. I don't think there is any difference with regard to security and functionality. Indeed, if it was automatically started then you are more free to just kill it whenever you want, which is better in some respects for security. > The idea of taking openssh and adding fsh-like abilities > and load-management abilities right into it is appealing. That would also be good. The code is pretty straightforward. Perhaps it would be hard to persuade the openbsd guys to take the patches, but it would be OK even if it existed as a nonstandard patch. -- Martin
pgp00000.pgp
Description: PGP signature
__ distcc mailing list http://distcc.samba.org/ To unsubscribe or change options: http://lists.samba.org/mailman/listinfo/distcc